Researchers at Purdue University, led by Vaneet Aggarwal and Dipesh Tamboli, have developed a patent-pending system aimed at enhancing privacy in generative AI used for photo editing. This innovative solution effectively masks sensitive attributes like eye color and facial hair before images are uploaded to AI platforms, significantly reducing the risk of identity exposure. Validated through rigorous testing, the system has shown an impressive 80% decrease in AI’s ability to detect these features, making it a critical advancement for secure online image sharing. Unlike traditional methods, which often compromise image quality or fail to protect biometric data, Purdue’s system maintains high-quality edits while ensuring full privacy by keeping sensitive data on the user’s device. This technology is compatible with existing generative AI models and aims to address pressing privacy concerns in a fast-evolving digital landscape. For industry partnerships interested in commercializing this technology, please contact Purdue Innovates Office of Technology Commercialization.
Source link
